Open beta: free while it lasts·Connect your first server in 10 minutes
Your on-call
SRE
Watches your servers while you sleep. When something breaks, it pings you on Telegram, explains what happened and helps you fix it.
One-command install · No credit card · Nothing runs without your approval
Works with what you already use
What it does for you
while you sleep
Fewer scares, answers in minutes and safe fixes, without hiring anyone or carrying a pager.
Monitoring · 3 servers
every 60 s- Disk (with forecast)3 serversOK
- Memory and OOM3 serversOK
- Sustained CPU3 serversOK
- Crash-looping containersshop-apiAttention
- Health checks14 containersOK
- TLS certificates5 domains21 days
- Server gone silent3 serversOK
Telegram · now
🔴 api-prod · shop-api is crash-looping (6 times in 10 min)
Deterministic rules: monitoring doesn't spend AI.
Monitoring · 3 servers
every 60 s- Disk (with forecast)3 serversOK
- Memory and OOM3 serversOK
- Sustained CPU3 serversOK
- Crash-looping containersshop-apiAttention
- Health checks14 containersOK
- TLS certificates5 domains21 days
- Server gone silent3 serversOK
Telegram · now
🔴 api-prod · shop-api is crash-looping (6 times in 10 min)
Deterministic rules: monitoring doesn't spend AI.
This is what 3 a.m. looks like
TopSentinel
bot
🔴 shop-prod · Disk will be full in ~40 min
Currently at 96%, growing 4 points an hour.
If you do nothing: the database will stop writing.
03:12
🔎 Diagnosis
System logs grew 3.1 GB since yesterday's deploy: the app was left in debug mode.
03:12
Only runs if you approve it
✅ Done · Journal shrunk to 500 MB (2.6 GB freed)
03:13
🔎 Check · Disk is now at 64%.
I'll remind you tomorrow to turn off debug mode.
03:14
From alert to fix, with your permission.
- Your serveragent · executorTopSentinelrules · AI · signerYouTelegram
$ curl -fsSL https://api.topsentinel.io/install.sh | sudo sh -s -- --token ts_••••
ts-agent installed · unprivileged
ts-executor installed · no network · actions disabled
Connected to TopSentinel (outbound, WSS) · no open ports
$ curl -fsSL https://api.topsentinel.io/install.sh | sudo sh -s -- --token ts_••••
ts-agent installed · unprivileged
ts-executor installed · no network · actions disabled
Connected to TopSentinel (outbound, WSS) · no open ports
Like having someone technical inside your servers.
Write to it like you'd write to a teammate. Before answering it looks at metrics, containers, logs, changes, ports, SSH logins and certificates.
Pick simple answers, no jargon, or technical ones, with numbers and commands. Try it in the chat.
TopSentinel
bot
How's everything?
Why does the API keep crashing?
What changed since yesterday?
What do I have exposed to the internet?
Limited hands, and always with your permission.
Letting something touch production takes trust. That's why the design assumes any piece can fail, ours included.
The AI never runs anything
It can only propose actions from a closed catalog. No free-form commands, no shell access.
Signed orders
Every approved action is signed with Ed25519. Your server verifies the signature, target, expiry (2 min) and that it was never used before.
No open ports
The agent only connects out to TopSentinel. The process with Docker access has no network.
TOTP for your data
Restarting a database or a queue asks for your 6-digit code, even if someone stole your Telegram.
You hold the switch
Actions ship disabled on every server. /pause blocks them across your whole account instantly.
Lightweight agent
Two ~6 MB Go binaries. About 18 MB of RAM between them. One-command install.
Built for you if…
You run 1 to 10 servers
On Hetzner, DigitalOcean, AWS or wherever, with Docker.
You deploy with compose
By hand, with Coolify or Dokploy, maybe with some help from AI.
You don't want to be on call
And you don't have, or need, an SRE team.
You'd rather get a Telegram than a complaint
Hear it from a message, not from an angry customer.
Running Kubernetes with hundreds of nodes and an on-call rotation? Then there are better-suited tools for you, and that's fine.
Sleep well.
Someone's on call.
We're in beta: free while it lasts. Connect your first server in under 10 minutes.